Interpreting the seasonal environmental history recorded by
1
Arctic bivalves
2
Vihtakari Mikko1,2,3,∗, Ambrose William G. Jr.2,4, Renaud Paul E.2,5, Locke William L. V4,
3
Carroll Michael L.2, Berge Jørgen1,5, Clarke Leon J.6, Cottier Finlo7, Hop Haakon3
4
1 Department of Arctic and Marine Biology, UiT The Arctic University of
5
Norway, N-9037 Tromsø, Norway
6
2 Akvaplan-niva, Fram Centre, N-9296 Tromsø, Norway
7
3 Norwegian Polar Institute, Fram Centre, N-9296 Tromsø, Norway
8
4 Department of Biology, Bates College, Lewiston, Maine 04240, USA
9
5 University Centre in Svalbard, N-9171 Longyearbyen, Norway
10
6 School of Science and the Environment, Faculty of Science and Engineering,
11
Manchester Metropolitan University, Manchester, M1 5GD, UK
12
7 Scottish Association for Marine Science, Scottish Marine Institute, Oban, Argyll
13
PA37 1QA, UK
14
∗E-mail: mikko.vihtakari@gmail.com
15
Keywords:Serripes groenlandicus;Ciliatocardium ciliatum; Laser-Ablation Inductively-Coupled-
16
Plasma Mass-Spectrometry; paleoclimatology; paleoceanography; bivalve mollusk shells;in situ
17
analyses; shell mineralogy
18
Abstract
19
Understanding rapid climate change in the Arctic and its ecosystem implications requires more
20
information on the environment at temporal resolutions and time-periods not available from
21
the instrumental records. Such information can be acquired through geochemical proxy records,
22
but sub-annual records are rare in the literature. We analyzed shell material of bivalve mol-
23
lusks (Serripes groenlandicus andCiliatocardium ciliatum) that were placed on oceanographic
24
moorings for one year in two Arctic fjords to assess the potential use of shell elemental ratios
25
as environmental proxies. Li/Ca, Mg/Ca, Li/Mg, Mn/Ca, Sr/Ca, Mo/Ca and Ba/Ca were de-
26
termined using Laser-Ablation Inductively-Coupled-Plasma Mass-Spectrometry. The mooring
27
exposure, combined with previously derived sub-annual shell growth models, allowed us to re-
28
late the elemental ratio patterns to oceanographic data (temperature, salinity, and fluorescence)
29
collected by instruments attached to the moorings. Shell Ba/Ca profiles were characterized by
30
abrupt peaks occurring 11 to 81 days after the phytoplankton bloom, as indicated by the sea-
31
water fluorescence index. Li/Ca and Mg/Ca values exhibited a logarithmic relationship with
32
shell growth rate, indicated by marginal R2 of 0.43 and 0.30, respectively. These ratios were
33
also linearly related to temperature, with marginal R2 of 0.15 and 0.17, respectively. Mn/Ca
34
and Sr/Ca ratios exhibited variability among individuals and their temporal pattern was likely
35
controlled by several unidentified factors. Mo/Ca patterns within the shells did not demon-
36
strate meaningful correlations with any mooring instrument data. Our results reflect complex
37
relationships between elemental ratios, bivalve metabolism, methodological limitations, and syn-
38
chronized environmental processes suggesting that none of the studied elemental ratios can be
39
used as all-encompassing proxies of seawater temperature, salinity, paleoproductivity, or shell
40
growth rate. Despite this, Ba/Ca and Li/Ca can likely be used as sub-annual temporal anchors
41
in further studies, as the deposition of these elements likely occurred simultaneously within each
42
fjord.
43
1 Introduction
44
The annual sea ice cover over the Arctic Ocean has declined by approximately 20 % since the
45
industrial revolution [data from Figure 4.3a in 1] with an accelerating rate over the last decade [2].
46
Such a reduction in sea-ice cover, together with other anthropogenic perturbations, is expected
47
to cause dramatic changes in Arctic marine ecosystems [2, 3]. Understanding and anticipating
48
these rapid changes requires information about the past climate at sufficient temporal resolution
49
and over longer time-periods than that usually provided by instrumental records [4]. Such
50
knowledge can be acquired by interpretation of geochemical proxy records, which can represent
51
long time scales [4–6]. Whereas records of environmental changes at longer than decadal time-
52
scales may indicate correlative relationships between climatic and biological patterns, combining
53
environmental and biotic data at sub-annual scales can help identify the ecological mechanisms
54
through which climate regulates biotic processes. Unfortunately, there are few sub-seasonal
55
high-resolution records presented in the literature due to a paucity of available data.
56
Shells of many filter-feeding bivalve mollusks are promising geochemical proxy archives due
57
to: 1) largely sedentary nature of bivalves, meaning that individuals record temporal rather
58
than spatial variability in seawater conditions; 2) distribution of bivalves across a wide variety
59
of habitats and latitudes [7]; 3) representation of bivalve shells in the geological record [7–10]; 4)
60
longevity of bivalves allowing longer than decadal proxy records per individual [11–13]; and 5)
61
regular growth patterns in bivalve shells that can be used to develop growth chronologies [14–
62
17]. Two common circumpolar bivalve species, the Greenland cockle (Serripes groenlandicus
63
Mohr, 1786) and the hairy cockle (Ciliatocardium ciliatum Fabricius, 1780), have been used
64
as environmental and climatic indicators in the previous studies [18–22]. They are long lived
65
species forming an aragonitic shell [23–25] with prominent annual growth lines deposited during
66
a slow winter shell growth period that is regulated by food availability [17, 19, 26]. Their shell
67
growth is further affected by temperature and often correlates with large scale climatic drivers
68
over annual to decadal scales [18, 20–22].
69
In theory, the environmental information stored in bivalve shells can be used to hind-cast sea-
70
water conditions with a sub-annual resolution based on geochemical proxies, such as element-to-
71
calcium ratios, that are sampled along chronologically deposited shell material [27–29]. Several
72
elemental ratios, such as Li/Ca [30, 31], Mg/Ca [32, 33], and Sr/Ca [34], have been suggested as
73
proxies of seawater temperature in bivalve shells, but these ratios are often affected by metabolic
74
and kinetic processes, and thus may be used as temperature proxies only for specific cases when
75
shell growth rate and seawater temperature are strongly intercorrelated [35, 36]. Lithium to
76
magnesium ratio could potentially be used to tease apart the metabolic effects in Li/Ca and
77
Mg/Ca [37]. The ratios of barium, manganese, molybdenum, and lithium to calcium have been
78
suggested as proxies of pelagic productivity [31, 38–40]. Barium to calcium provides one of the
79
most consistent elemental ratio signals in bivalve shells: Ba/Ca profiles are characterized by a
80
flat background signal that is periodically interrupted by sharp peaks in a wide range of species
81
across various habitats and latitudes [24, 38, 39, 41–48]. In addition to potentially representing
82
variability in primary productivity, Ba/Ca may indicate ambient seawater concentrations [49].
83
In contrast, manganese is often associated with shell precipitation rate and may also be influ-
84
enced by seawater redox conditions, and therefore shows variable patterns depending on species
85
[50–54]. Molybdenum, on the other hand, may be incorporated through diet, making Mo/Ca a
86
potential proxy of paleoproductivity [40, 49].
87
Consequently, the development of elemental ratios in bivalve shells as environmental proxies
88
could be valuable, especially in the Arctic where instrumental records are short or interrupted
89
and climate change is rapid [55]. Elemental ratio proxies in bivalve shells are, however, compli-
90
cated by metabolism as calcium carbonate mineralization does not occur directly from seawater,
91
but takes place in a chemically controlled space; the extrapallial cavity [56–58]. Interpretation
92
of these geochemical proxies is further complicated by shell growth rate, which varies through
93
the year [17] and appears to influence some element ratios [36]. Consequently, understanding
94
the sub-annual growth patterns is a fundamental prerequisite for using any shell-based proxy
95
at sub-annual resolution. Very few studies, and none in the Arctic, have been able to relate
96
elemental ratios measured within bivalve shells to seawater parameters data recorded at the
97
growth location with sub-annual resolution.
98
In this study, we examine minor and trace elemental ratios within the shells ofS. groen-
99
landicusandC. ciliatum, and assess their potential use as environmental proxies. We deployed
100
these bivalves on moorings in two oceanographically contrasting fjords in Svalbard for one year
101
[17, 26]. The bivalve deployment combined with previously obtained sub-annual growth models
102
[17] allowed us to relate the elemental ratio patterns to the oceanographic data recorded by
103
mooring instrumentation. We aimed to examine whether: 1) Li/Ca, Ba/Ca, Mn/Ca or Mo/Ca
104
could be used as proxies of primary productivity as has been suggested by other studies, 2)
105
Li/Ca, Mg/Ca, Li/Mg or Sr/Ca could be used as proxies of temperature or shell growth rate,
106
and 3) any of the above mentioned elemental ratios were deposited simultaneously in different
107
individuals indicating that they could be used as sub-annual chronological markers in the studied
108
species.
109
2 Materials and Methods
110
2.1 Study design
111
A suite of element (Li, Mg, Mn, Sr, Mo, and Ba) to calcium ratios was determined for sub-annual
112
patterns in shells of two bivalve species (Serripes groenlandicus andCiliatocardium ciliatum)
113
deployed on oceanographic moorings for one year during the periods September 2007–2008 and
114
September 2009–2010 in two fjords on Svalbard: Kongsfjorden and Rijpfjorden. These two fjords
115
are oceanographically different. Kongsfjorden is an Atlantic water-influenced open fjord, whereas
116
Rijpfjorden is a fjord with a sill (depth 100-200 m) that is influenced mainly by Arctic water
117
masses [59–62]. Kongsfjorden was ice-free throughout the field deployment with the exception
118
of occasional drift ice, whereas Rijpfjorden was covered by sea ice for 8 months (January 21–
119
September 16) in 2007–2008 [63], and for 5 months (February 15–July 21) in 2009-2010 [17]. The
120
bivalve deployment on moorings is described in detail by Ambrose Jret al.[26] and Vihtakari
121
et al.[17]. In brief, bivalves were collected from the western Barents Sea in August 2007 and
122
from Svalbardbanken in August 2009. They were held in flow-through seawater tanks for 1–
123
4 weeks at the University Centre in Svalbard and incubated in seawater with 125 mg L−1 of
124
calcein dye for 24 h immediately before they were placed in 7 mm mesh plastic cages (hereafter
125
baskets) on the oceanographic moorings. The calcein mark was used as an absolute time marker
126
of deployment and was identified in sectioned shells using fluorescent imaging [see 17]. During
127
2009-2010, the bivalves were deployed to two water depths, 15 m (basket A) and 25 m (basket
128
B), while in 2007-2008 they were deployed only to 25 m (Table 1). The bivalves were deployed
129
in September each year and recovered one year later.
130
Bivalves collected from the moorings were sacrificed and shells then were embedded in epoxy
131
resin [as described in 26]. Embedded shells were cut into thick sections along the maximum
132
growth axis, as described in Vihtakari et al. [17], and the thick sections were polished to a
133
thickness of 2.0±0.1 mm. These thick sections then were transferred to a clean room, where
134
they were rinsed and brushed in Milli-Q water, sonicated for 5 min and rinsed again. Finally,
135
the thick sections were left to dry overnight before they were analyzed using Laser-Ablation
136
Inductively-Coupled-Plasma Mass-Spectrometry (hereafter LA-ICP-MS). Eleven shells were fur-
137
ther analyzed forin situδ18O values using secondary ion mass spectrometry (SIMS) to determine
138
sub-annual growth models [see 17]. Measured element ratio patterns determined for nine shells
139
that demonstrated adequate growth models were compared to weekly averages of seawater tem-
140
perature, salinity and fluorescence index records obtained from mooring instruments located
141
adjacent to bivalve baskets (Table 2, see 17 for details).
142
2.2 Elemental ratio analyses
143
LA-ICP-MS [64] was conducted at the Plasma Mass Spectrometry Facility, Woods Hole Oceano-
144
graphic Institute (MA, US), using a Thermo-Finnigan Element2 HR-ICP-MS coupled to a New
145
Wave Laser UP 193 nm excimer laser ablation system. A sequence of holes was ablated along
146
the middle of the shell thick section from the outer margin to the calcein line [see 17] using
147
95 s dwell time, 10 Hz repetition rate and 90% output power. The analysis was conducted in
148
2009 for 2007-2008 deployment specimens and in 2011 for 2009-2010 deployment individuals.
149
Magnesium (25Mg), calcium (48Ca), manganese (55Mn), strontium (88Sr) and barium (138Ba)
150
were analyzed in both years. Molybdenum (98Mo) and lithium (7Li) were added to the analysis
151
for 2009-2010 samples. Due to the low concentration of Mo in the CaCO3matrix, 2009-2010
152
shells had a larger ablation crater size [¯x= 87.5±0.7µm (SE), n = 612] compared to 2007-2008
153
samples [¯x = 42.0± 0.3µm (SE), n = 311]. The distance between laser holes [¯x = 104.1±
154
14.3 (SD)µm ] was kept constant between sessions and samples, and therefore the number of
155
ablation holes varied between 17 and 64 per analyzed shell depending on the length of annual
156
growth increment.
157
The signal intensity (counts per second) of the analyzed elements was monitored in an
158
Element2 low resolution mode during the LA-ICP-MS analyses. The recording of element signal
159
intensity was started approximately 10 s after initiating the laser ablation to clean the shell
160
surface of debris and to ensure that the ablation plume material had reached the ICP-MS. An
161
estimated value for each element was generated by averaging 50 signal intensity measurements
162
during the peak of material flow. Nitric acid (5 % HNO3) was used as a blank, ensuring a
163
constant flow of the acid into the ICP-MS. Every tenth sample analyzed was a blank. The
164
moving average of blanks was calculated and subtracted from the data. Since the analyzed
165
shell matrix was predominantly aragonite [23, 25], 48Ca was used as an internal standard by
166
normalizing all other elements to Ca concentration [65]. Two standards, Japanese Certified
167
Reference Material or “JpnCRM” [66] and FEBS-1 [67], were run as every tenth and twentieth
168
sample, respectively. These standards were used to correct for instrument drift and to calibrate
169
elemental ratios to cover all isotopes. FEBS-1 was used for Mn/Ca and Li/Ca and JpnCRM
170
for the other elemental ratios. The reference materials did not have a certified value for Mo.
171
Therefore, Mo/Ca concentrations are given as percentage of Mo/Ca maximum for each shell
172
and comparison of absolute Mo/Ca values was not possible
173
2.3 Datasets and statistical analyses
174
The position of the LA-ICP-MS holes was related to sub-annual growth lines and a measurement
175
axis that was related to the historical location of the shell margin using ImageJ [68] and sclero
176
package [69] for R software [70], as described in Vihtakariet al.[17]. The method also allowed
177
a spatial estimation of averaging error [71, 72]. Resulting LA-ICP-MS sample distances are
178
therefore expressed as mm from deployment (i.e. the calcein mark) along the measurement axis,
179
together with minimum and maximum extents for each LA-ICP-MS hole (Figures S1–S6).
180
Growth models for nine shells (three from each basket: KB, RA and RB, Table 1), based on
181
estimated daily growth trajectories for SIMSδ18O centroids (Figure 9 in 17), allowed comparison
182
of elemental ratio data to mooring instrument data (temperature, fluorescence index and salinity)
183
and modeled growth rate. The estimated temporal extent sampled by each LA-ICP-MS hole was
184
used to calculate average growth rate, temperature, salinity, and fluorescence index values that
185
were used as predictor variables in consequent regression models. The averages were calculated
186
using daily values. The relationship between element ratios (response variable in all models)
187
and shell growth rate was logarithmic, and therefore growth rates were log-transformed before
188
analyses.
189
Linear mixed-effect regression models (LMMs) were used to examine the overall relationships
190
in the dataset by using samples as random effects, assuming a random intercept and a constant
191
slope (see Table S3 and Text S1 for definitions of the models). In order to examine the overall
192
variance of each elemental ratio explained by each predictor variable, LMMs were run separately
193
with each non-transformed predictor variable (Model 1; Table S3). Marginal and conditional
194
R2values for LMMs for these models were calculated using MuMIn package [73] for R [70] and
195
the method described by Johnson [74]. Marginal R2 values were used as a measure of overall
196
variance explained by each response variable and to examine whether the proxy relationship was
197
constant among samples. To examine the overall relative importance of each predictor variable
198
and the direction of the linear relationship, all predictor variables were combined as fixed effects
199
into a same LMM (Model 2; Table S3). Response variables were log-transformed, and predictor
200
variables centered to their means and scaled to their standard deviations before running Model
201
2. The fixed effects (effects of each predictor variable to an elemental ratio) then were scaled
202
to the maximum absolute value of 95% confidence intervals resulting to a measure of relative
203
effect for each fixed effect. Linear mixed-effect models were calculated using the nlme package
204
[75]. The variability in relationships between response and predictor variables among individual
205
samples was examined using linear regression models fitted for each sample, response variable
206
and predictor variable separately (Model 3; Table S3).
207
Coefficients of variation (CV) for minimum and maximum elemental ratios over the mooring
208
deployment were used to assess among individual consistency of elemental ratios using all ana-
209
lyzed shells over two deployment periods (n = 30, Table 1). Correlations between elemental ratios
210
and predictor variables for regression models were examined using principal component analy-
211
sis [76] calculated on correlation matrices averaged over samples using Fisher z-transformation
212
[77–79]. These correlation matrices are presented in Table S4.
213
3 Results
214
3.1 Oceanographic conditions in the fjords
215
Kongsfjorden experienced warmer temperatures in 2007-2008 than in 2009-2010 (Figure 1): The
216
autumn (September to December) temperatures in Kongsfjorden were on average 1.0◦C higher
217
in 2007 compared to 2009, the winter (January to April) temperatures 1.7◦C warmer, and the
218
spring/summer (May to September) temperatures 2.6◦C warmer in 2008 compared to 2010. In
219
contrast, temperature differences between years varied in Rijpfjorden: The autumn (September
220
to November) temperatures in Rijpfjorden were also on average 1.0◦C higher in 2007 compared
221
to 2009, the winter (December to May) temperatures were almost equal between deployment
222
years, but the summer temperatures were on average 2.4◦C lower in 2008 compared to 2010. In
223
Kongsfjorden, temperature began to increase in May in both years. In 2007-2008, temperature
224
remained above zero, while in the winter of 2009-2010, temperature was generally below zero.
225
Temperature was recorded at two depths (15 and 25 m) in 2009-2010. Temperature differences
226
between depths were generally small, except during the summer stratification period, when
227
temperature at 15 m was approximately 1◦C higher than at 25 m. Rijpfjorden experienced
228
temperatures close to -1.7◦C from January until July (6 months) in 2007-2008 and from Jan-
229
uary until June (5 months) in 2009-2010. Temperature rose abruptly in mid-July 2010, whereas
230
in 2008 it started increasing in mid-May, but did not exceed 0 ◦C. In 2009-2010, tempera-
231
tures were similar at both measured depths until late August, when the surface layer cooled by
232
approximately 3◦C relative to the deeper (25 m) layer.
233
In both fjords, the fluorescence index (FLI) was close to zero prior to a dramatic increase
234
during the spring (Figure 1). The first fluorescence peak occurred later (mid-June to mid-July)
235
in Rijpfjorden than in Kongsfjorden (mid-May to beginning of June). Salinity was relatively
236
stable in Kongsfjorden, with a range between 33.3 and 35.0 (Figure 1). Rijpfjorden experienced
237
variable salinity regime, related to melt water from sea ice, from July to December. Salinity
238
varied more in 2009-2010 (34.6-30.6) than in 2007-2008 (34.3-31.7), and was most variable at
239
the shallow baskets (15 m).
240
3.2 Patterns in element ratio profiles
241
Lithium to calcium ratios were consistently lower during winter and increased after the winter
242
growth band in all studied shells (Figures 2, S3–S6). The increase occurred simultaneously
243
with increased growth rate in growth modeled shells (Figures 2 and S7). Minimum Li/Ca was
244
13.9±0.3 (SE, n = 22)µmol mol−1on average (Table 3). The Li/Ca minimum was deposited
245
sometime between October and late May in Kongsfjorden and between October and mid-July
246
in Rijpfjorden (Figure 2). Coefficient of variation for minimum Li/Ca values varied between
247
7.5 and 14.1 % among baskets and was higher than that for maximum values (Table 3). The
248
maximum values were 21.6±0.3 (SE, n = 22) on average, and were estimated to occur July to
249
early September in Kongsfjorden and mid-July to early August in Rijpfjorden (Figure 2).
250
Magnesium to calcium ratios were at their lowest during the winter growth band and in-
251
creased immediately after or towards the end of the winter growth period in most analyzed
252
shells (Figures 2 and S1–S6). Three shells deployed to Rijpfjorden in 2007, however, did not
253
demonstrate clear seasonal Mg/Ca fluctuations (Figure S2). The strongest increase in Mg/Ca
254
values occurred during spring together with increased growth rate (Figures 2 and S8). After
255
reaching the maximum in July to mid-August in Kongsfjorden and in late July to late August in
256
Rijpfjorden, Mg/Ca values decreased slightly until the end of the deployment period (Figure 2).
257
Maximum Mg/Ca values ranged between 1.04 and 4.15 mmol mol−1 being generally higher in
258
2009-2010 than in 2007-2008 (Table 3). Minimum Mg/Ca values ranged between 0.39 and 1.70
259
mmol mol−1and were not obviously different among years. Coefficient of variation for Mg/Ca
260
minimum and maximum values was higher than that for Li/Ca (Table 3).
261
Manganese to calcium values exhibited variable patterns, but were also characterized by
262
peaks deposited during the translucent summer growth period in 24 of 30 analyzed shells (Figure
263
S1–S6). These peaks were deposited sometime between late May and August in Kongsfjorden,
264
and between early July and early August in Rijpfjorden occurring one to 70 days after the
265
fluorescence peak (Table 4 and Figure 2). Low Mn/Ca values were deposited during the winter
266
growth band from January until the end of the winter growth band (Figure S9). Average
267
maximum manganese values ranged between 1.31 and 8.52µmol mol−1 (Table 3). Maximum
268
Mn/Ca values within baskets showed high variability as illustrated by coefficient of variation
269
(Table 3). Average minimum Mn/Ca values ranged between 0.16 and 0.75µmol mol−1among
270
baskets, and coefficient of variation was high (Table 3). Average minimum and maximum values
271
were lower in 2009-2010 (Table 3).
272
Individuals within baskets demonstrated considerable variability with respect to Sr/Ca pro-
273
files (Figures 2, S1–S6). Minimum values were deposited before the winter growth band in 3
274
samples, during the winter growth in 4 samples, and after the winter growth in 23 samples.
275
Furthermore, maximum Sr/Ca values occurred before, during and after the winter growth band
276
in 7, 7, and 16 samples, respectively (Figures S1–S6). Minimum Sr/Ca values were deposited
277
between May and August in two growth modeledS. groenlandicus from Kongsfjorden and be-
278
tween October and March in the growth modeledC. ciliatumspecimen (Figures 2 and S10). In
279
Rijpfjorden, the minimum values were deposited between July and mid-August in seven shells
280
and between April and mid-July in oneS. groenlandicusspecimen (Figure 2). Maximum Sr/Ca
281
values in growth modeled shells from Kongsfjorden were deposited at the end of the mooring de-
282
ployment in mid-September, whereas Rijpfjorden shells showed more variability with maximum
283
values occurring in the beginning of the mooring deployment (September to December) as well
284
as towards the end of the mooring deployment (August to September, Figure 2). Coefficient
285
of variation for minimum and maximum Sr/Ca values was lower than those for Mg/Ca (Table
286
3). Minimum Sr/Ca value was 1.32±0.04 mmol mol−1(SE, n = 30) on average and maximum
287
value 2.37±0.09 mmol mol−1(SE, n = 30).
288
Molybdenum to calcium ratios were at their highest during or before the winter growth
289
band in all shells analyzed for Mo/Ca (2009-2010) and the ratios decreased after the end of
290
the growth check (Figures 2, S3–S6). After the minimum Mo/Ca, which occurred between
291
mid-April and September in Kongsfjorden and between July and August in Rijpfjorden, Mo/Ca
292
values increased again until the end of the mooring exposure (mid-September 2010, Figures 2
293
and S11). Maximum Mo/Ca values were measured at the beginning of the mooring deployment
294
(September to April, Figure 2).
295
Barium to calcium profiles were characterized by abrupt unimodal peaks (maximum values
296
= 3.1-76.1µmol mol−1, ¯x= 20.0µmol mol−1, Table 5) that were differentiated from low Ba/Ca
297
background levels (0.43–2µmol mol−1, ¯x= 1µmol mol−1, Figures 2, S1–S6). The peaks appeared
298
annually, occurring subsequent to the winter growth band in 27 of 30 analyzed shells (Figures
299
S1–S6). A distinct barium peak was not present in twoC. ciliatum from 2007-2008 deployed
300
in the 25 m basket in Rijpfjorden (Figure S2) and oneC. ciliatum from 2009-2010 deployed
301
in the 15 m basket in Kongsfjorden (Figure S3). In 2009-2010 samples, the Ba maxima were
302
considerably lower in the 25 m basket in Rijpfjorden compared to other baskets (RB in Table 3
303
and Figure S6). Barium peak values were not consistent within a basket as indicated by high
304
coefficient of variation (Table 3). The minimum Ba/Ca values were associated with a lower
305
within basket variability than the maximum values (Table 3). Barium peaks in Kongsfjorden
306
were estimated to occur between June and mid-August, 18 to 100 days after the fluorescence
307
peak (Table 5). Further, Ba/Ca peak values were deposited in July in Rijpfjorden occurring 11
308
to 36 days after the first peak in fluorescence index (Table 5).
309
3.3 Correlations between element ratios, growth rates and mooring instru-
310
ment data
311
Li/Ca and Mg/Ca covaried within 2009-2010 shells as indicated by arrows pointing approxi-
312
mately to the same direction in the PCA plot (Figure 3B) and high correlation coefficients (rz 313
= 0.78, r = 0.13–0.92; Table S4). Similar correlations between element ratios were evident for
314
Sr/Ca and Mo/Ca in 2009-2010 (Figure 3B,rz= 0.59, r =−0.69–0.99), Mn/Ca and Ba/Ca –
315
especially in the growth modeled shells (Figure 3C,rz= 0.50, r =−0.02–0.78), and Mg/Ca and
316
Mn/Ca in 2007-2008 shells (Figure 3A,rz= 0.38, r =−0.31–0.78). Further, Mg/Ca and Li/Mg
317
were strongly negatively correlated in 2009-2010 shells as demonstrated by arrows pointing to
318
opposite directions in the PCA plot (Figure 3B,rz=−0.92, r =−0.99 –−0.60). Also Li/Ca
319
and Li/Mg, Mg/Ca and Mo/Ca, and Li/Ca and Mo/Ca were negatively correlated (Table S4).
320
Temperature and salinity were negatively correlated (rz=−0.71, r =−0.86–0.57), whereas tem-
321
perature yielded positive correlations with fluorescence (rz= 0.48, r = 0.34–0.67) and logarithm
322
of shell growth rate (rz= 0.43, r = 0.22–0.66, Figure 3D).
323
Overall, logarithm of growth rate was the best explanatory factor for element ratio variability
324
in growth modeled shells (Figure 4A). Coefficient of determination (R2) for individual samples
325
ranged between 0.19 and 0.75 for the regression between Li/Ca and growth rate, between 0.30
326
and 0.59 for Mg/Ca, between 0.11 and 0.24 for Li/MG, and between 0.01 and 0.87 for Mn/Ca
327
(Table S2). Also Sr/Ca exhibited significant regressions with growth rate, but these relation-
328
ships varied from positive to negative (Table S2). Temperature yielded significant regressions
329
with Li/Ca, Mg/Ca, Li/Mg, and Sr/Ca (Table S2), but in the majority of samples these regres-
330
sions were not as strong as those for logarithm of growth rate (Figure 4A). The temperature
331
relationships for Li/Ca, Mg/Ca and Li/Mg were relatively consistent among samples, although
332
associated with large residual standard error (Tables S1–S2).
333
4 Discussion
334
Barium, manganese, molybdenum, and lithium to calcium ratios have previously been related to
335
primary production [31, 38–40] (Section 4.1). Although Mn/Ca and Ba/Ca exhibited patterns
336
that resembled the patterns in the fluorescence index (Figure 2), which was used as a proxy
337
of primary production, the differences in peak heights among samples from the same basket
338
suggested that these element ratios were also affected by other processes and could not be used
339
as straightforward proxies of primary production (Tables 3–5; see Section 4.1). Despite this,
340
Ba peaks were deposited likely at the same in a basket, but the timing varied between baskets
341
occurring 11 to 81 days after the phytoplankton bloom (Figure 2, Table 5, Section 4.3). Ba/Ca
342
could potentially be related to dissolved or particular Ba in ambient seawater. Mo/Ca and
343
Li/Ca did not exhibit patterns that could have been linked to primary production (Figure 2).
344
Lithium, magnesium and strontium to calcium ratios, in turn, have been suggested as proxies
345
of growth rate or temperature [30–34] (Section 4.2). We did observe considerable similarities
346
between Li/Ca, Mg/Ca, growth rate and temperature (Figures 3–4), but individual samples
347
from a same basket demonstrated variability in element-to-calcium ratios making it difficult to
348
use these ratios as proxies of absolute growth rate or temperature (see Section 4.2). Neverthe-
349
less, Li/Ca might reflect crystal growth rate in bivalve shells, whereas Mg/Ca appears to be
350
loosely linked with temperature (Figures 5–6). Finally, individuals within baskets demonstrated
351
variability in Sr/Ca profiles that could not satisfactorily be explained by any single predictor
352
variable (growth rate, temperature, fluorescence and salinity) used in this study (Figure 4).
353
In general, our results highlight the limitations caused by metabolically controlled deposition
354
of CaCO3in bivalves [56, 57] suggesting that none of the studied element ratio could be used
355
as straightforward proxies of temperature, salinity, paleoproductivity or shell growth rate. In
356
following sections we discuss the studied element ratios as potential proxies of primary produc-
357
tion (Section 4.1), shell growth rate or temperature (Section 4.2), and sub-seasonal temporal
358
anchors (Section 4.3). We also highlight the methodological constraints associated with our data
359
(Section 4.4).
360
4.1 Potential proxies of primary production
361
Barium to calcium profiles were characterized by distinct unimodal peaks, which resembled the
362
peaks in fluorescence index (Figures 1, 2, and S1-S6). The barium peak in Kongsfjorden shells
363
occurred approximately 74 days after the peak in phytoplankton bloom, which took place in
364
mid-May, and 19 days after ice-algae/phytoplankton associated fluorescence peak in Rijpfjor-
365
den (Table 5). Dissolved barium from seawater, which in turn is sometimes connected with
366
phytoplankton blooms [46, 80], has been found to consistently incorporate into calciticMytilus
367
edulisandPecten maximusshells with a partition coefficient of approximately 0.1 [39, 49]. Ap-
368
plied to our shells, Ba/Ca values should have been approximately similar, within the averaging
369
error framework (see Section 4.4), in each basket assuming that calcium was uniformly dis-
370
tributed along studied shells. Measured Ba/Ca background values varied between 0.4 and∼2
371
µmol mol−1, were consistent with those reported earlier [46], and did not show any obvious vari-
372
ation within baskets that could not have been explained by averaging error (Table 3). Measured
373
maximum Ba/Ca values, on the other hand, varied between 3.1–76.1µmol mol−1demonstrating
374
different peak values among shells from a same basket (Table 3). This variability in maximum
375
values is among the largest reported [46], and cannot completely be explained by averaging error
376
(see Section 4.4).
377
Predictor variables did not satisfactorily explain the Ba/Ca peaks: although Ba/Ca peaks
378
occurred simultaneously with increased shell growth in all growth modeled shells (Figures 2
379
and S12), growth rate explained only 2% of Ba/Ca variation across samples (marginal R2from
380
LMM; Figure 4) and<1 to 18% among samples (R2from regression models; Table S2). Further,
381
temperature was negatively related with Ba/Ca explaining 2% of variation across samples (Fig-
382
ure 4). Bivalve age, shell height, or length of the growth increment during mooring deployment
383
did not yield significant slopes in a regression model with Ba/Ca peak values, but Ba/Ca peak
384
values were significantly lower in the 25 m basket in Rijpfjorden compared to other baskets.
385
Therefore, our results are inconclusive about the environmental factors associated with the ob-
386
served barium peaks. Nevertheless, the considerable differences in Ba/Ca maximums among
387
samples from a same basket and the variable time-lag from bloom between fjords (Table 5) sug-
388
gest that although Ba/Ca might be connected to processes related to primary production, the
389
ratio cannot be used as a direct paleoproductivity proxy, agreeing with what has been suggested
390
by recent studies [45, 46, 49, 81].
391
In addition to barium, manganese to calcium profiles also demonstrated peaks that resembled
392
the fluorescence index peaks (Figures 1, 2 and S1-S6). Despite the seemingly synchronous
393
deposition of Mn in growth modeled shells (Table 4), Mn/Ca patterns exhibited individual
394
differences among shells from a same basket (Table 3). Further, Mn/Ca values in the growth
395
modeled shells were clearly correlated with growth rate (Figure 4 and Table S2) demonstrating
396
that Mn/Ca incorporation is likely, at least partly, kinetically controlled. Manganese occurs
397
partly as non-lattice-bound element in an aragonitic bivalveCorbula amurensis[82]. A varying
398
amount of Mn not directly bound to CaCO3matrix could also explain the mixed Mn/Ca results
399
in our study. Nevertheless, Mn/Ca peaks occurring approximately simultaneously in growth
400
modeled shells also demonstrate a degree of synchronous environmental or physiological control.
401
Previous studies suggest that Mn/Ca could partly be incorporated in relationship with Mn
402
concentration in seawater [51, 83]. Phytoplankton blooms have also been suggested as a cause
403
for Mn fluctuations in bivalve shells [24, 38]. Our data do not support the direct connection with
404
phytoplankton bloom events, but it is possible that pelagic Mn cycle is connected to productivity
405
to some extent as reviewed by [83]. Consequently, Mn/Ca is a potential, but complicated proxy
406
of several environmental and physiological factors in both species.
407
Maximum molybdenum to calcium values were measured during autumn before the depo-
408
sition of the winter growth band in all growth modeled shells (Figure 2). Consequently, our
409
dataset did not demonstrate prominent Mo peaks occurring during spring as has been reported
410
for calcitic scallopsComptopallium radula [48] andP. maximus[40]. Nevertheless, Mo/Ca pro-
411
files were relatively similar among shells demonstrating that Mo/Ca values either fell under the
412
detection limit of ICP-MS or that the incorporation mechanism could have been environmentally
413
regulated. The incorporation of Mo into bivalve shells might occur through diet, which makes
414
Mo/Ca a promising environmental proxy [40, 49]. If this was the case local phytoplankton may
415
not have been enriched in Mo. Alternatively, Mo could be connected to sediment surface redox-
416
processes [28] or sediment particles, as bivalves in our study were deployed in the water column
417
and did not grow in their natural habitat. Although our results do not preclude the possibility
418
for Mo/Ca being a potential proxy inS. groenlandicusandC. ciliatum, more research is needed
419
to draw further conclusions about this elemental ratio.
420
Our data did not demonstrate a clear connection between fluorescence index and Li/Ca
421
(Figures 4, and S7) casting a doubt on the hypothesis of phytoplankton blooms causing Li/Ca
422
peaks [31]. Therefore, Li/Ca peaks cannot be used as a proxy of timing and magnitude of
423
phytoplankton blooms in studied shells, although it is possible that phytoplankton blooms could
424
have contributed to increasing the Li/Ca values in Kongsfjorden (Table S2).
425
4.2 Potential proxies of growth rate or temperature
426
Lithium to calcium patterns were similar among individuals in baskets suggesting synchronized
427
responses to environmental or physiological processes (Figure 2 and Table 3). Logarithm of
428
average growth rate explained 43% of overall Li/Ca variation across all samples (LMM, Figure 4),
429
and 19–75% among samples (regressions, Table S2). Li/Ca–shell growth rate relationships were
430
logarithmic unlike in previous published studies where the authors reported linear relationships
431
with a similar slope forP. maximus[31] andArctica islandica[30] (Figure 5A). Shell growth rate
432
is an indicator of crystal growth rate in bivalve mollusk shells [31, 84]. Therefore, the positive
433
correlations between Li/Ca and shell growth rate agree with other published studies suggesting
434
that crystal growth rate is likely the primary driver of Li/Ca incorporation in bivalve mollusk
435
shells [30, 31]. Nevertheless, studies report differing regression equations between Li/Ca and
436
shell growth rate and these relationships do not yield particularly high R2values (Figure 5A).
437
This suggests that also other factors affect Li/Ca incorporation.
438
Temperature and riverine output have also been suggested to partly control Li/Ca in bivalve
439
shells [30, 31]. Since temperature and growth rate were correlated in our shells [17], the effects of
440
these factors are difficult to separate. Nevertheless, temperature significantly explained Li/Ca
441
variability, although these correlations were generally not as strong as for shell growth rate (Fig-
442
ures 4–5 and Table S2). The imprecision in our growth models could have contributed to the
443
lower temperature correlations, as a one-month shift in Li/Ca peak would have led to consider-
444
ably stronger temperature correlations for Rijpfjorden shells (Figures 2 and S7). Despite this,
445
the relationships for species that have been studied so far do not appear to demonstrate strong
446
enough R2values to reconstruct seawater temperatures (Figure 5B). Instead, significant regres-
447
sions between Li/Ca and temperature in bivalve mollusk shells (Figure 5B) could be explained
448
by dependency between temperature and shell growth rate, and therefore CaCO3crystal growth
449
rate.
450
Since we lack element concentration measurements in seawater, we can only speculate about
451
the effect of riverine output increasing Li concentration in ambient water and therefore contribut-
452
ing to shell Li/Ca [30]. Li/Ca peaks were coincident with decreased salinity (Figures 2 and S7).
453
If melt-water events increased Li concentration in ambient water in our study, it is possible that
454
these events could have contributed to Li/Ca fluctuations as suggested by Th´ebaultet al.[30].
455
Despite the uncertainties in our dataset, we can conclude, with a relatively high certainty, that
456
Li/Ca cannot be used as a temperature proxy inS. groenlandicus andC. ciliatumshells, but
457
appears to be a promising proxy of shell and/or crystal growth rate. Li/Ca, however, did not
458
yield strong enough relationships to precisely reconstruct sub-annual shell growth.
459
Relatively consistent patterns in Mg/Ca among individuals from the same basket (Figures 2,
460
S1–S6) suggested that the incorporation of Mg/Ca is likely related to synchronized environmental
461
or physiological processes. A large coefficient of variation, however, indicates that these processes
462
do not yield similar Mg/Ca peak values among shells (Table 3). Relatively strong correlations
463
with logarithm of average growth rate indicated that incorporation of Mg/Ca could be related to
464
shell precipitation rate similarly to Li/Ca (Figure 4). Furthermore, Mg/Ca correlated positively
465
with temperature (Figure 4 and Table S2). Many studies have reported similar significant
466
correlations between Mg/Ca ratio and sea surface temperature [32–35, 44, 53, 85–88]. Most of
467
these studies report either a large variability in temperature correlations similar to our study
468
[e.g. 34, 35], or that the relationship is restricted to certain conditions [e.g. 86, 87]. Organic
469
matter prior the elemental analysis has been removed in some studies that have reported strong
470
relationships between temperature and Mg/Ca [32, 89].
471
Our Mg/Ca–temperature relationships are similar to those reported for calcitic bivalves
472
Mytilus trossulus[32],M. edulis[90], andP. maximus[88] with the exception that coefficients of
473
variation are clearly lower in our study (Figure 6). Mg/Ca is thought to be strongly metabolically
474
controlled in marine bivalves: present day Mg/Ca molar ratio is 5.2 mol mol−1[91], but report
475
Mg/Ca ratios in bivalve CaCO3 that are several orders of magnitude lower than the ambient
476
molar ratios (varied between 0.0041 and 0.0004 mol mol−1in this study). Furthermore, Mg/Ca
477
is precipitated to inorganic aragonite following an inverse relationship with expected molar
478
ratio of>0.085 mol mol−1 for the temperatures in this study [92]. Despite this, most reported
479
Mg/Ca–temperature relationships are positive (Figure 6),Crassostrea gigasbeing an exception
480
[54]. It should also be noted that Mg/Ca–temperature relationships appear generally stronger
481
for calcitic bivalves (bivalves in Figure 6) than for aragonitic bivalves (such asS. groenlandicus,
482
C. ciliatumandA. islandica[e.g. 35]). It seems feasible that Mg/Ca functions as a temperature
483
proxy in many bivalve shells (Figure 6), but Mg/Ca incorporation is also influenced by other
484
factors such that the imprecision associated with temperature estimates derived from Mg/Ca
485
is often larger than the seasonal temperature fluctuations. Our results are consistent with this
486
hypothesis and indicate that Mg/Ca is an unreliable temperature proxy forS. groenlandicus
487
andC. ciliatum. Nevertheless, our results also indicate that temperature does correlate with
488
Mg incorporation, and further studies should consider removal of organic matter before ICP-MS
489
analyses.
490
Studies on corals have demonstrated that combining Li/Ca and Mg/Ca could potentially
491
be used to tease apart the metabolic effects associated with these ratios and strengthen the
492
temperature relationship [37]. Our results, however, demonstrated generally weaker correlations
493
between Li/Mg and temperature than those between Li/Ca and temperature and Mg/Ca and
494
temperature separately (Figure 4, Table S2). Consequently, Li/Mg does not provide a robust
495
temperature proxy.
496
Strontium-to-calcium ratio was significantly affected by all predictor variables (Figure 4),
497
temperature and fluorescence index yielding the most consistent regressions (Table S2). Coeffi-
498
cient of variation for Sr/Ca maximum values indicates that Sr/Ca values varied among samples
499
from a same basket (Table 3). The large variability in Sr/Ca among samples from a same
500
location is consistent with the literature [44, 93] and suggests that any environmental signals
501
in Sr/Ca may be difficult to separate from vital effects. Strontium partition into calcium car-
502
bonate is related to the crystal growth rate of CaCO3matrix [92, 94]. Although, some earlier
503
studies have successfully used Sr/Ca as a temperature proxy [85, 95, 96], more recent studies
504
question the relationship [50, 97, 98]: it seems possible that temperature and crystal growth
505
rate of CaCO3 skeleton are connected resulting in a positive correlation between Sr/Ca and
506
temperature. Judging from our data, this was not the case for studied shells.
507
4.3 Sub-seasonal temporal anchors
508
Barium-to-calcium maximum values were deposited at approximately same time among samples
509
from the same basket (Table 5) considering the uncertainty caused by LA-ICP-MS averaging
510
error and growth models derived fromδ18O values (see Section 4.4). Measured Ba/Ca maximums
511
were estimated to be deposited in mid-July to early August in Kongsfjorden (Table 5). Barium
512
peaks in Rijpfjorden occurred during or right after a fast shell growth period (Figures 2 and S12)
513
and were timed to occur early July in the basket at 15 m depth and late July, 12 days later, in
514
the deeper basket at 25 m depth (Table 5). Simultaneous occurrence of Ba/Ca maximums within
515
baskets and similar patterns in 29 of 32 analyzed shells (Figure S1-S6) indicates synchronous
516
environmental or physiological drivers for incorporation of Ba in studied shells. Synchronously
517
deposited chemical proxies are useful temporal anchors to combine chronologies across bivalves
518
sampled from the same location [29]. Our results indicate that the Ba/Ca peaks are likely to
519
occur simultaneously 2.5 months to 2.5 weeks after primary production bloom, and they can be
520
used as sub-annual anchors across shells from a same location, if averaging error of elemental
521
sampling is kept sufficiently low.
522
Li/Ca also demonstrated remarkably synchronous patterns within baskets (Table 3) as min-
523
imum and maximum value variability could likely be explained by averaging error caused by
524
LA-ICP-MS sampling (see Section 4.4). Therefore, Li/Ca peak and trough values could have
525
been approximately similar across individuals from a same basket further demonstrating the
526
synchronized incorporation of this element ratio. Overall, Li/Ca ratios corresponded with those
527
reported by Th´ebaultet al.[30]: the range of Li/Ca fluctuation they reported was 1.3 to 1.6
528
fold over a growing season, whereas lithium values in this study varied between 1.3 and 2.2 fold
529
(1.6 on average). This demonstrates that Li/Ca could work as a temporal anchor also for other
530
species thanS. groenlandicus andC. ciliatum. Since Li/Ca peaks were rather broad in studied
531
shells it is advisable to use the increases in Li/Ca as temporal anchors.
532
4.4 Methodological limitations
533
The bivalves in this study were held in the water column on oceanographic moorings, and
534
therefore they might not have recorded elemental ratios similarly to their natural habitat. The
535
mooring deployment likely excluded the effect of sediment-surface redox-processes, which have
536
been suggested as important contributors for the seasonal dynamics of, at least, Mn [28, 48,
537
83, 99]. Further, we did not observe similar seasonal patterns in Sr/Ca ratios that has been
538
reported earlier forS. groenlandicus[18, 24]. It is possible that Sr/Ca is partly connected with
539
sediment surface processes and therefore our shells did not record all possible variability for this
540
element ratio.
541
The extent of time averaging sampled by LA-ICP-MS is relative to the sample volume and
542
average shell growth rate over the sampled area [71, 72]. Because sample hole size in our study
543
varied little within years (see Section 2.2), time averaging was related to shell growth rate. Even
544
though LA-ICP-MS sampling was able to capture the Ba/Ca peaks (Figures 2, S1–S6) it is
545
possible that time-averaging contributed to profiles of some elements during low growth rate
546
such that no meaningful environmental correlations were found [100].
547
Growth models used to determine the time extent for each LA-ICP-MS sample were subject
548
to uncertainty [17]. It is unlikely that these growth models were an entirely accurate representa-
549
tion of the actual growth during the mooring deployment and, therefore, our dataset contained a
550
bias, which increased correlations between element ratios and average shell growth rate, because
551
shell growth rate was obtained from growth models, which affected the alignment of elemental
552
ratios. Further, shell growth rate and temperature were significantly correlated in all growth
553
modeled shells (Figure 3; 17).
554
Even though we attempted to keep LA-ICP-MS samples as close to the middle of the shell
555
section as possible, non-linear growth patterns could have caused variations in the actual location
556
of LA-ICP-MS samples hence affecting the element ratios [101], since the sample spot alignment
557
method used in this study [69] could not correct for measurement bias caused by variability in
558
CaCO3matrix. Furthermore, the sample alignment method assumed two-dimensional sampling
559
ignoring any effects of LA-ICP-MS sample volume. Consequently, the curvature of growth
560
lines deeper in the sample could have increased imprecision of element ratios through three-
561
dimensional time averaging. Despite all these uncertainties, our dataset is extensive and clearly
562
indicates that all of the studied elemental ratios were affected by several factors to the extent that
563
no element ratio in this study could be used as an absolute straightforward proxy of temperature,
564
salinity, fluorescence or shell growth rate.
565
5 Conclusions
566
We conclude that Ba/Ca, Li/Ca and Mg/Ca have a potential as environmental proxies inS.
567
groenlandicusandC. ciliatumshells: Incorporation of Ba/Ca might be connected with seasonal
568
dissolved or particular Ba dynamics in ambient water, and incorporation of Li/Ca and Mg/Ca
569
are likely connected with both CaCO3crystal growth rate and seawater temperature. Despite
570
this, all studied element ratios were likely affected by multiple internal and external factors
571
complicating the interpretation of element ratios. Our study was further affected by method-
572
ological constraints, such as time-averaging error, experimental artifacts, and uncertainties in
573
sub-annual growth models leading to partly inconclusive results for Sr/Ca and Mo/Ca. Despite
574
this our results are an important contribution to high-latitude bivalve shell geochemisty high-
575
lighting that none of the studied elemental ratios can be used as all-encompassing proxies of
576
seawater temperature, salinity, paleoproductivity, or shell growth rate. This, however, does not
577
preclude the use of element-to-calcium ratios as environmental proxies, but merely indicates that
578
seasonal dynamics of elements in seawater and seasonal variations in bivalve metabolism must
579
be understood better to link the elemental ratios in bivalve mollusk shells with environmental
580
processes.
581
Acknowledgments
582
We acknowledge the use of the NSF-supported WHOI ICP-MS facility and thank Scot Birdwhis-
583
tell for his excellent assistance. We are grateful to Bates Imaging Center and William Ash for
584
help with thick-section photographs. Further, we want to thank the Stack Exchange community
585
for help with the graphical presentation and data-analysis, and the R community for maintaining
586
open source statistics tools used in this study. This research was financed through the UiT The
587
Arctic University of Norway Utenlandstipend (MV), the EU 7th Framework Program project
588
Arctic Tipping Points (contract number FP7-ENV-2009-226248; http://www.eu-atp.org; PER),
589
the Research Council of Norway project Havet og Kysten (184719/S40; PER), the Norwegian
590
Polar Institute (HH, MV) and Akvaplan-niva (PER, MV, WGA, MLC).
591
References
592
1. Vaughan, D.; Comiso, J.; Allison, I.; Carrasco, J.; Kaser, G.; Kwok, R.; Mote, P.;
593
Murray, T.; Paul, F.; Ren, J.; Rignot, E.; Solomina, O.; Steffen, K.; Zhang, T. Ob-
594
servations: Cryosphere. InClimate Change 2013: The Physical Science Basis Contri-
595
bution of Working Group I to the Fifth Assessment Report of the Intergovernmental
596
Panel on Climate Change; Stocker, T.; Qin, D.; Plattner, G.K.; Tignor, M.; Allen, S.;
597
Boschung, J.; Nauels, A.; Xia, Y.; Bex, V.; Midgley, P., Eds.; Cambridge University
598
Press: Cambridge, UK and New York, NY, USA, 2013; chapter 4, pp. 317–382.
599
2. AMAP. Arctic climate issues 2011: Changes in Arctic snow, water, ice and permafrost.
600
SWIPA 2011 Overview report. Technical report, Arctic Monitoring and Assessment
601
Programme (AMAP), Oslo, Norway, 2012.
602
3. Grebmeier, J.M. Shifting patterns of life in the Pacific Arctic and sub-Arctic seas.Ann
603
Rev Mar Sci2012,4, 63–78.
604
4. Masson-Delmotte, V.; Schulz, M.; Abe-Ouchi, A.; Beer, J.; Ganopolski, A.; Rouco,
605
J.G.; Jansen, E.; Lambeck, K.; Luterbacher, J.; Naish, T.; Osborn, T.; Otto-Bliesner,
606
B.; Quinn, T.; Ramesh, R.; Rojas, M.; Shao, X.; Timmermann, A. Information from
607
paleoclimate archives. InClimate Change 2013: The Physical Science Basis Contri-
608
bution of Working Group I to the Fifth Assessment Report of the Intergovernmental
609
Panel on Climate Change; Stocker, T.; Qin, D.; Plattner, G.K.; Tignor, M.; Allen, S.;
610
Boschung, J.; Nauels, A.; Xia, Y.; Bex, V.; Midgley, P., Eds.; Cambridge University
611
Press: Cambridge, UK and New York, NY, USA, 2013; chapter 5, pp. 383–464.
612
5. Pearson, P.N.; Palmer, M.R. Atmospheric carbon dioxide concentrations over the past
613
60 million years. Nature2000,406, 695–699.
614
6. Zachos, J.; Pagani, M.; Sloan, L.; Thomas, E.; Billups, K. Trends, rhythms, and
615
aberrations in global climate 65 Ma to present.Science2001,292, 686–93.
616
7. Dame, R.F. Introduction. In Ecology of Marine Bivalves, An Ecosystem Approach;
617
CRC Press: Boca Raton, FL, USA, 2012; pp. 1–19.
618